Table Of Content

ContemporaryMathematics 1 1 The Homotopy Theory of Function Spaces: A Survey 0 2 n Samuel Bruce Smith a J Abstract. Wesurveyresearchonthehomotopytheoryofthespacemap(X,Y) 3 consistingofallcontinuousfunctionsbetweentwotopologicalspaces. Wesum- 1 marizeprogressonvariousclassificationproblemsforthehomotopytypesrep- resented bythepath-components ofmap(X,Y). Wealsodiscussworkonthe ] T homotopy theory of the monoid of self-equivalences aut(X) and of the free loop space LX. We consider these topics in both ordinary homotopy theory A aswellasafterlocalization. Inthelattercase,wediscussalgebraicmodelsfor . thelocalizationoffunctionspaces andtheirapplications. h t a m [ 1. Introduction. 2 Inthispaper,wesurveyresearchinhomotopytheoryonfunctionspacestreated v as topological spaces of interest in their own right. We begin, in this section, 4 0 with some general remarks on the topology of function spaces. We then give a 8 brief historical sketch of work on the homotopy theory of function spaces. This 0 sketch serves to introduce the basic themes around which the body of the paper is . 9 organized. 0 By work of Brown [39, 1964] and Steenrod [260, 1967], the homotopy the- 0 ory of function spaces may be studied in the “convenient category” of compactly 1 generated Hausdorff spaces. Retopologizing is required, however. Given spaces X : v and Y in this category, let YX denote the space of all continuous functions with i X the compact-open topology. Define r map(X,Y)=k(YX) a to be the associated compactly generated space. Then map(X,Y) satisfies the desired exponential laws and is a homotopy invariant of X and Y. The space map(X,Y)isgenerallydisconnectedwithpath-componentscorrespondingtotheset of free homotopy classes of maps. We write map(X,Y;f) for the path-component containing a given map f: X →Y. Important special cases include: map(X,Y;0), the space of null-homotopic maps; map(X,X;1), the identity component; aut(X), 2010 Mathematics Subject Classification. 55P15, 55P35, 55P48, 55P50, 55P60, 55P62, 55Q52,55R35,46M20. Key words and phrases. Function space, monoid of self-equivalence, free loop space, space ofholomorphicmaps,gaugegroup,stringtopology,configurationspace,sectionspace,classifying space,Gottliebgroup,localization,rationalhomotopytheory. (cid:13)c0000(copyrightholder) 1 2 SAMUELBRUCESMITH the space of all homotopy self-equivalences of X; and LX = map(S1,X) the free loop space. Concrete results on the path-components of map(X,Y) often require much more restrictive hypotheses on X and Y. By Milnor [214, 1959], when X is a compact, metric space and Y is a CW complex, the components map(X,Y;f) are of CW homotopy type. A natural case to consider then is when X is a finite CW complex and Y is any CW complex. By Kahn [153, 1984], map(X,Y) is also of CW type when X is any CW complex and Y has finitely many homotopy groups. The space map(X,Y) has two close relatives. If X and Y are pointed spaces, we havemap (X,Y) the spaceof basepointpreservingfunctions, with components ∗ map (X,Y;f) for f a based map. Given a fibration p: E → X, we have Γ(p) ∗ the space of sections with components Γ(p;s) for s a fixed section. Of course, map(X,Y) ≃ Γ(p) when p fibre-homotopy trivial with fibre Y. Many theorems aboutmap(X,Y)generalizetoΓ(p)andmanyhaverelatedversionsformap (X,Y). ∗ Forthe sakeofbrevity,when possiblewe statetheoremsfor the free functionspace andomitextensionsandrestrictions. Theoremsstatedforthebasedfunctionspace are then results that do not apply to map(X,Y). 1.1. A Brief History. Function spaces are at the foundations of homotopy theory and appear in the literature dating back, at least, to Hurewicz’s definition of the homotopy groups in the 1930s. Work focusing explicitly on the homotopy theory of a function space first appears in the 1940s. Whitehead [286, 1946] introducedthe problemofclassifyingthe homotopytypes representedbythe path- components of a function space, focusing on the case map(S2,S2). Hu [148, 1946] showed π1(map(S2,S2;ιm))∼=Z/2|m|, where ι is the map of degree m thus distinguishing components of different abso- m lute degree. Adecadelater,papersofThom[270,1957]andFederer [88,1956]appeared givingdualmethodsforcomputinghomotopygroupsofcomponentsofmap(X,Y). Thom used a Postnikov decomposition of Y to indicate a method of calculation. Federerconstructedaspectralsequenceconvergingtothesehomotopygroupsusing a cellular decomposition of X. Both authors obtained the following basic identity: πq(map(X,K(π,n);0))∼=Hn−q(X;π) for X a CW complex and π an abelian group. Inthe1960s,themonoidaut(X)ofallhomotopyself-equivalencesofX emerged asacentralobjectforthetheoryoffibrations. Stasheff[259,1963]constructeda universalfibrationforCWfibrationswithfibreofthehomotopytypeofafixedfinite CW complex X, building on work of Dold-Lashof [72, 1957]. His result implied theuniversalX-fibrationisobtained,uptohomotopy,byapplyingtheDold-Lashof classifying space functor to the evaluation fibration ω: map(X,X;1) → X. In this same period, Gottlieb [108, 1965] introduced and studied the evaluation subgroups or Gottlieb groups: G (X)=im{ω : π (map(X,X;1))→π (X)}⊆π (X) n ♯ n n n initiatingarichliteratureontheevaluationmap. Amongmanyotherproperties,he showedthe Gottlieb groups correspondto the image of the linking homomorphism in the long exact sequence of homotopy groups of the universal X-fibration. Thus THE HOMOTOPY THEORY OF FUNCTION SPACES: A SURVEY 3 thevanishingofaGottliebgroupG (X)isequivalenttothevanishingofthelinking n homomorphism in degree n for every CW fibration with fibre X. In the 1970s, Hansen [129, 1974] began a systematic study of the homotopy classification problem for the path components of map(X,Y). He completed the classification for map(Sn,Sn) building on the methods of Whitehead, mentioned above. He and other authors obtained complete results in many special cases in- volving spheres, suspensions, projective spaces and certain manifolds. The space of holomorphic maps Hol(M,N) between two complex manifolds offersadeeprefinementofthehomotopyclassificationproblemforcontinuousmaps withimportantinterdisciplinaryconnections. Segal[244,1979]initiatedthestudy of the space Hol(M,N) in homotopy theory proving the inclusion Hol∗(S2,CPm)֒→map (S2,CPm;ι ) k ∗ k induces a homology equivalence through a range of degrees. Here Hol∗(S2,CPm) k denotes the space of based holomorphic maps of degree k. In fundamental work incomplex geometry, Gromov [116, 1989]identified the classof elliptic manifolds and proved they satisfy the “Oka Principle”. As a consequence, he identified a largeclassofmanifoldsforwhichtheinclusionHol(M,N)֒→map(M,N)isaweak equivalence. Cohen-Cohen-Mann-Milgram [56, 1991] described the full stable homotopytype of Hol∗(S2,CPm), their descriptiongivenin terms of configuration k spaces. A related problem of stabilization for moduli spaces of connections is the subjectofthe famous“Atiyah-Jonesconjecture” inmathematicalphysicsAtiyah- Jones [15, 1978]. The gauge groups provide a connection between the homotopy theory of function spaces and the theory of principal bundles. Let P: E → X be a principal G-bundle for G a connected topological group classified by a map h: X → BG. The gauge group G(P) of P is defined to be the group of G-equivariant homeo- morphisms f: E → E over X. Atiyah-Bott [14, 1983] used the gauge group in their celebrated study of Yang-Mills equations and principal bundles over a Rie- mann surface. They made use of Thom’s theory and a multiplicative equivalence originally due to Gottlieb [112, 1972] G(P)≃Ωmap(X,BG;h) to study the homotopy theory of BG(P). Gottlieb’s identity, in turn, links the classification of gauge groups up to H-homotopy type, for fixed G and X, to the homotopyclassificationproblemformap(X,BG). Crabb-Sutherland[67,2000] provedthatthegaugegroupsG(P)representonlyfinitelymanyhomotopytypesfor G a compact Lie groupand X a finite complex,. In contrast,the path-components ofmap(X,BG) may representinfinitely many distinct homotopytypes inthis case by Masbaum [199, 1991]. The advent of localization techniques introduced new depth to the study of function spaceswhile opening up a wide rangeof fundamentalproblems andappli- cations. In his seminal paper on rational homotopy theory, Sullivan [263, 1977] sketcheda constructionforanalgebraicmodelforcomponentsofmap(X,Y)forX and Y simply connected CW complexes with X finite, as an extension of Thom’s ideas. Sullivan’s construction was completed by Haefliger [123, 1982]. Sullivan also identified the rational Samelson Lie algebra of aut(X) for X a finite, simply connected CW complex via an isomorphism: π∗(aut(X))⊗Q,[ , ] ∼= H∗(Der(MX)),[ , ] 4 SAMUELBRUCESMITH Here the latter space is the homology of the Lie algebra of degree lowering deriva- tions of the Sullivan minimal model of X with the commutator bracket. One of the early applications of Sullivan’s rational homotopy theory was the proofbyVigué-Poirrier-Sullivan[282,1976]oftheunboundednessoftheBetti numbersofthefreeloopspaceLX =map(S1,X)forcertainsimplyconnectedCW complexes X. Combined with a famous result of Gromoll-Meyer [115, 1969]in geometry,thiscalculationsolvedthe“closedgeodesicproblem”formanymanifolds. The calculation was deduced from a Sullivan model constructed for LX. The p-localhomotopy theory of a function space featured in a landmark result in algebraic topology, the proof of the Sullivan conjecture. Miller [213, 1984] proved π (map (Bπ,X;0))=0 for all n≥0 n ∗ where π is any finite group and X any finite CW complex. Among many applications, this result was used by McGibbon-Neisendorfer [209, 1984] to affirm Serre’s conjecture: π (X) contains a subgroup of order p for infinitely many m. m Lannes [180, 1987]constructed the T-functor which is left adjoint to the ten- sor product in the category of unstable modules over the Steenrod algebras. His constructionprovidedamodelforthemodpcohomologyofthe spacemap(BV,X) where V is a p-group. Lannes’ construction was adapted to the rational homotopy settingbyBousfield-Peterson-Smith [30,1989]and,later,Brown-Szczarba [37, 1997] to give another model for the rational homotopy type of map(X,Y;f). Fresse [102, preprint]recently constructeda versionof Lannes’functor in a category of operadic algebras giving a model for the integral homotopy type of certain function spaces. The free loop space recently re-emerged as a central object for study in homotopy theory with the appearance of work of Chas-Sullivan [50, preprint]. They constructed a product on the regraded homology H (LMm)=H (LMm) ∗ ∗+m for a simply connected, closed, orientedm-manifold Mm using intersection theory. They also defined a bracket on the equivariant homology of LMm and a degree +1 operator giving H (LMm) the structure of Batalin-Vilkovisky algebra. These ∗ structureshaveincarnationsindiverseothersettings. Theirstudy,knownasstring topology, is now an active subfield in the intersection of homotopy theory and geometry. 1.2. Organization. In Section 2, we discuss workon the ordinaryand stable homotopy theory, as opposed to the local homotopy theory of function spaces. We focus on the areas introduced above, namely: (i) the general path component map(X,Y;f); (ii) the monoid aut(X); and (iii) the free loop space LX. We also discussworkonthestablehomotopytheoryofthesespacesandonspectralsequence calculationsoftheirinvariants. InSection3,we discussthe localizationoffunction spaces. We describe the algebraic models of Sullivan, and of later authors, for the general component, the monoid of self-equivalences and the free loop space in rational homotopy theory, and survey their applications. We also discuss the p- local homotopy theory of function spaces including the work of Miller, Lannes and others on the space of maps out of a classifying space, and algebraic models for function spaces in tame homotopy theory. The paper includes a rather extensive THE HOMOTOPY THEORY OF FUNCTION SPACES: A SURVEY 5 bibliography gathering together both papers directly focused on function spaces and papers giving significant applications and extensions. 2. Ordinary and Stable Homotopy Theory of Function Spaces. We divide our discussionin this sectionaccording to the cases (i), (ii) and (iii) above. Wethendiscusssomegeneralresultsinstablehomotopytheoryandspectral sequence constructions for function spaces. 2.1. General Components. Asmentionedintheintroduction,thefollowing open problem lies at the historical roots of the study of function spaces as objects in their own right. Problem 2.1. Given spaces X and Y classify the path-components map(X,Y;f) up to homotopy type for homotopy classes f: X →Y. We consideravarietyofcasesherebeginning withthe mostclassical,mention- ing progress on Problem 2.1, when appropriate. 2.1.1. Maps from Spheres and Suspensions. The components of map(Sp,Y) correspondtothehomotopyclassesinπ (Y). ThecoproductonSp givesrisetoan p equivalencemap (Sp,Y;α)≃map (Sp,Y;0),foranyclassα: Sp →Y. Byadjoint- ∗ ∗ ness, πn(map∗(Sp,Y;0)) ∼= πn+p(Y). These observations were made by White- head[286,1946]whogavethefirstalgebraicmethodforcomputation. Whitehead identified the linking homomorphism in the long exact homotopy sequence for the evaluation fibration map (Sp,Y;α)→map(Sp,Y;α)→Y obtaining: ∗ ··· // π (Y) ∂ // π (map (Sp,Y;α)) //π (map(Sp,Y;α)) //··· n+1 QQQWQQ(QαQ)QnQQQQ(( ∗ (cid:15)(cid:15)∼= n π (Y) n+p where W(α)(β) = −[α,β] denotes the Whitehead product map. Using this se- w quence,heprovedmap(S2,S2;ι)6≃map(S2,S2;0)bycomparinghomotopygroups. Hu[148,1946]andKoh[160,1960]computedπ (map(S2m,S2m;α))forsmall 2m−1 values of m. In these cases, the order of π (map(S2m,S2m;α)) depends on the 2m−1 absolute value of the degree of α and so distinguishes components with different absolute order. Since clearly map(S2m,S2m;α)≃ map(S2m,S2m;−α) the classification in these cases was complete with these calculations. Hansen [129, 1974] obtained the complete classification for self-maps of Sn. For even spheres, he proved map(S2m,S2m;α)≃map(S2m,S2m;β) ⇐⇒ [α,ι] =±[β,ι] . w w Here ι ∈ π (S2m) is the fundamental class. For odd spheres, the components of 2m map(S2m−1,S2m−1)areallhomotopyequivalentform=1,2,4duetotheexistence of a multiplication on S2m−1 in these cases. For m 6= 1,2,4, Hansen showed map(S2m−1,S2m−1;ι)6≃map(S2m−1,S2m−1;0) and map(S2m−1,S2m−1;ι) if deg(α)=odd map(S2m−1,S2m−1;α)≃ map(S2m−1,S2m−1;0) if deg(α)=even. (cid:26) Problem 2.1 remains open for map(Sm,Sn) for m > n. Yoon [305, 1995] observeda connection between the Gottlieb group G (Y) and the homotopy clas- m sificationproblem for map(Sm,Y) showing map(Sm,Y;α)≃map(Sm,Y;0) if and 6 SAMUELBRUCESMITH only if α ∈ G (Y). Lupton-Smith [193, 2008] extended this to a surjection of m sets {components map(Sm,Y;f)} π (Y)/G (Y) //// . m m homotopy equivalence Thus the complexity of the classificationproblem for map(Sm,Sn) is roughly that of computing Gottlieb groups G (Sn). Extensive, low-dimensional calculations of m this groupwererecently made by Golasin´ski-Mukai[106, 2009]. Lee-Mimura- Woo [184, 2004] calculated the Gottlieb groups for certain homogeneous spaces. WhenX =ΣAis asuspension,the fibres map (ΣA,Y;f)ofthe variousevalu- ∗ ation fibrations ω : map(ΣA,Y;f)→ Y are all homotopy equivalent to the space f map (ΣA,Y;0) with homotopy groups ∗ π (map (ΣA,Y;0))=[Σq+1A,Y]. q ∗ Lang [178, 1973] extended Whitehead’s exact sequence to this case replacing the Whitehead product in π (Y) by the generalized Whitehead product in [Σ∗A,Y]. ∗ It is natural to consider, as Whitehead did, a stronger version of Problem 2.1, namely, the classification of the evaluation fibrations ω : map(X,Y;f) → Y up f to fibre homotopy type for homotopy classes f: X → Y. Hansen [128, 1974] defined ω : map(ΣA,ΣB;f) → Y to be strongly fibre homotopy equivalent to f ω : map(ΣA,ΣB;g) → ΣB if the fibre homotopy equivalence is homotopic to the g identity after (fixed) identification of the fibres with map (ΣA,ΣB;0). He proved: ∗ ω is strongly fibre homotopic to ω ⇐⇒ [f,1 ]=[g,1 ] f g ΣB ΣB where[, ]heredenotesthe generalizedWhiteheadproductin[ΣA,ΣB]. McClen- don [206, 1981] showed that the evaluation fibrations ω : map(ΣA,Y;f) → Y f behave as principal fibrations and, in particular, are classified by maps s: Y → map(A,Y) determined by generalized Whitehead products. 2.1.2. Maps into Eilenberg-MacLane Spaces. The weak homotopy type of the space map(X,K(π,n);f) may be described for any f: X →K(π,n) for π abelian. The ideas are due to Thom [270, 1957] with a refinement by Haefliger [123, 1982]. First, observe that these components are all homotopy equivalent since K(π,n) has the homotopy type of a topological group. A homotopy class α ∈ π (map(X,K(π,n;0)))corresponds,byadjointness,toamapA: Sp×X →K(π,n). p On cohomology, A∗(x )=1⊗a +u ⊗a n n p n−p where a ,a ∈H∗(X;π) with subscripts indicating degree while u ∈Hp(Sp;π) n n−p p and x ∈ Hn(K(π,n);π) are the fundamental classes. Since A restricts to the n constant map on Sp × ∗ we see a = 0. The assignment α 7→ a gives the n n−p identification πp(map(X,K(π,n);f))∼=Hn−p(X;π), mentioned in the introduction and leads to directly to a weak equivalence map(X,K(π,n);f))≃ K(Hn−p(X;π),p). w p≥1 Y Thom also indicated how the homotopy groups π (map(X,Y;f)) for Y a finite p Postnikovpiece are determined, up to extensions, by the k-invariantsof Y and the groups Hn−p(X,π (Y)). This approachwas encoded in Haefliger’s construction of n a Sullivan model for map(X,Y;f), as discussed below. THE HOMOTOPY THEORY OF FUNCTION SPACES: A SURVEY 7 Gottlieb [110, 1969] extended Thom’s result to the case n = 1 and π any group. Here map(X,K(π,1);f)≃ K(C(f ),1) w ♯ where C(f ) denotes the centralizer of the image of f : π (X) → π. Møller ♯ ♯ 1 [218, 1987] showed that when Y is a twisted Eilenberg-Mac Lane space, then map(X,Y;f) is one also with homotopy groups determined by the cohomology groups of X with twisted coefficients in the homotopy groups of Y. Note that the weak equivalences above are homotopy equivalences by Whitehead’s Theorem, when map(X,K(π,n)) is of CW type, e.g., when X is compact or a CW complex. In general, the study of the homotopy type of map(X,Y;f) when Y has at least two nonvanishing homotopy groups is a difficult, open problem. 2.1.3. Maps between Manifolds. The homotopy theory of map(Mm,Nn) for Mm and Nn closed manifolds is a topic of wide-ranging interest. In this case, important variations have been considered. Below we consider one such variation with direct ties to Problem 2.1, namely spaces of holomorphic maps. We begin with the space map(Mm,Nn). If T is an orientable surface, then T ≃ K(π (T ),1) and the classification g g 1 g problem for map(X,T ) reduces to the computation of centralizers of homomor- g phismsintoπ (T ). Forg ≥2thisgroupishighlynonabelianandtheonlypossible 1 g nontrivial centralizers are isomorphic to Z by Hansen [135, 1983]. Hansen [131, 1974]earlierconsideredthe space map(T ,S2). As a generalizationofWhitehead’s g exact sequence, he showed an exact sequence 0→Z/2|m|→π (map(T ,S2;ι ))→Z2g →0 1 g m which gives the classification, in terms of degree, in this case. The fundamental groupπ (map(T ,S2;ι ))waslatercompletelydeterminedbyLarmore-Thomas 1 g m [182, 1980]. Hansen [133, 1981] extended his classification result for the space of self- maps of spheres to the case map(Mm,Sm) where Mm is closed, oriented, con- nectedmanifold with vanishing firstBetti number. Note that, by Hopf’s Theorem, [Mm,Sm] = Z with maps classified by degree. When m is even and ≥ 4, Hansen showed the homotopy types of map(Mm,Sm;α) are classified by the absolute val- uesofthe degreesoftheα. Whenmisoddandm6=1,4,7therearetwohomotopy types corresponding to the distinct types map(Mm,Sm;0) and map(Mm,Sm;ι) where ι is of degree 1. Again in this case, components are classified by the parity of degree of the class α. Sutherland [264, 1983] extended Hansen’s work eliminating the restriction onthefirstBettinumberanddealingwiththecaseMmnonorientable. Notethatin thelattercasethereareonlytwodistinctclassesα: Mm →Sm andsotheproblem reduces to distinguishing these components for m 6= 1,4,7. Sutherland observed that the components of map(Mm,Sm) allhave the same homotopy type if there is a map I: Mm →map(Sm,Sm;ι) making the diagram map(Sm,Sm;1) pppppIpppppp77 (cid:15)(cid:15)ω Mm ι //Sm 8 SAMUELBRUCESMITH commute, where ι is of degree 1. Taking Mm = RPm, we have a lift I based on the lift I′: RPm → SO(n+1) ⊆ map(Sm,Sm;1) of ι. Sutherland showed the components of map(Mm,Sm) are all of the same homotopy type if there exists a map f: Mm →RPm of odd degree giving examples with m 6= 1,4,7 for which all the components are homotopy equivalent. Sasao[235,1974]studiedthehomotopytypeofcomponentsofmap(CPm,CPn;i) for m≤n and i: CPm →CPn the inclusion. He constructed a map α : U(n+1)/∆(m+1)×U(n−m)→map(CPm,CPn;i) m,n where∆(m+1)⊂U(m+1)denotesscalarmultipliesoftheidentity. Heprovedα m,n induces an isomorphism on rational homotopy groups and on ordinary homotopy groups through degree 4n−4m+1. Yamaguchi [289, 1983] extended Sasao’s analysistothecaseofquaternionicprojectivespaces. Møller[215,1984]gavethe completeclassificationforthecomponentsmap(CPm,CPn)showingthehomotopy typesareclassifiedbytheabsolutevalueofthedegreeofarepresentativeclass. The result is a direct consequence of his calculation n+1 H2n−2m+1(map(CPm,CPn;ιk))∼=Z/d where d= |k|m. m (cid:18) (cid:19) Yamaguchi [292, 2006] studied maps between real projective spaces. He defined the analogue of Sasao’s map, here of the form β : O(n+1)/∆(m+1)×O(n−m)→map(RPm,RPn;i) m,n and proved β is an equivalence on ordinary and rational homotopy groups m,n through certain ranges of degrees. When G is a topological group (or group-like space) the path-components of map(X,G) are all of the same homotopy type. Problem 2.1 thus reduces, in this case, to the study of the homotopy theory of the null-component map(X,G;0). Given Lie groups G and H, the calculation of homotopy invariants of map(G,H) is a difficult open problem. Recently, Maruyama-O¯shima [198, 2008] computed the homotopy groups of map (G,G) for G=SU(3),Sp(2) in degrees ≤8. ∗ 2.1.4. SpacesofHolomorphicMaps. Segal[244,1979]provedabasicresulton thehomotopytheoryofthe spaceHol(M,N). Hisworklaunchedavitalsubfieldof researchonthe“stability”oftheinclusionHol(M,N)֒→map(M,N).Segalproved Hol∗(T ,CPn)֒→map (T ,;ι ) k g ∗ g k induces a homology isomorphism through dimension (k−2g)(2n−1) where T is g a Riemann surface of genus g. Specializing to the case of the sphere, he proved Hol∗(S2,CPn)֒→map (S2,CPn;ι ) k ∗ k induces a homotopy equivalence up to degree 2n−1. Segal’s work was extended by many authors. Guest [119, 1984] proved the corresponding stability result on homology for Hol∗(S2,F) ֒→ map(S2,F;ι ) for k k certain complex flag manifolds F. His proof involved developing the analogue of a Morse-theoreticresultforthecaseoftheenergyfunctionalonthespaceC∞(S2,F) ofsmoothmaps. Kirwan[158,1986]extendedSegal’sresulttothecasethetarget is the complex Grassmannian manifold G(n,n +m) of n-planes in n+ m-space proving Hol∗(S2,G(n,n + m)) ֒→ map(S2,G(n,n + m);ι ) induces a homology k k isomorphism in degrees depending on k,n and m. Mann-Milgram [196, 1991] THE HOMOTOPY THEORY OF FUNCTION SPACES: A SURVEY 9 consideredthiscaseaswell,constructingaspectralsequencetoanalyzethehomol- ogy of Hol∗(S2,G(n,n+m)). Gravesen [113, 1989] studied holomorphic maps k into space ΩG for G a complex, compact Lie group. Cohen-Cohen-Mann-Milgram[56,1991]andCohen-Shimamota[64,1991] described the stable homotopy type of Hol∗(S2,CPn). They proved k Hol∗(S2,CPn)≃C (R2,S2n−1) k k where C (R2,S2n−1) is the configurationspace of distinct points in R2 with labels k in S2n−1 of length at most k. Cohen-Cohen-Mann-Milgram also computed the homologyofHol∗(S2,CPn)withZ -coefficientsintermsofDyer-Lashofoperations. k p Mann-Milgram[197]usedthestablehomotopydecompositionabovetoprovethe homologystabilityoftheinclusionHol∗(S2,F)֒→map (S2,F)forF anSL(n,C)- k ∗ flag-manifold. Boyer-Mann-Hurtubise-Milgram [33, 1994] and Hurtubuise [149, 1996] proved homology stabilization theorems for the space Hol∗(S2,G/P) k forcertaincomplexhomogeneousspacesG/P. Boyer-Hurtubise-Milgramgave a configuration space description of Hol (T ,M) for certain complex manifolds k g admitting nice Lie group actions extending the approach of Gravesen. Segal’s stabilization problem has deep interdisciplinary connections. Gromov [116, 1989] obtained general stability results as a consequence of his work on the Oka Principle in complex geometry. A complex manifold M satisfies the Oka principle if every continuous map f: S → M is homotopic to a holomorphic map whereS isaSteinmanifold. Gromovidentifiedtheclassof“elliptic”manifoldsand provedellipticmanifoldssatisfiedtheOkaprinciple. Consequently,heobtainedthe inclusion Hol(S,M)֒→map(S,M) is a weak homotopy equivalence for S Stein and M elliptic. The class of elliptic manifolds includes complex Lie groups and their homogeneous spaces. TheproblemofstabilizationalsohasafamousincarnationinYang-Millstheory and mathematical physics. Atiyah-Jones [15, 1978] constructed a map θ : M →map (S3,SU(2);ι ) k k ∗ k where M is a moduli space of connections on a principal SU(2)-bundle P over k k S4 corresponding to a map S4 → BSU(2) of degree k. They proved θ induces k a homology surjection through a range of degrees and conjectured θ induces an k equivalence in both homotopy and homology through a range depending on k. Work on the Atiyah-Jones conjecture includes Taubes [268, 1989], Gravesen [113, 1989]and Boyer-Hurtubise–Mann-Milgram [31, 1993]. Many authors have studied related spaces of maps. Vassiliev [277, 1992] proved a stable equivalence Hol∗(S2,CPn)≃SPk (C) k n−1 where the latter is the space of monic complex polynomials of degree k with all roots of multiplicity < n. Guest-Kozlowski-Yamaguchi [122, 1994] extended Segal’s result in a different direction, proving the inclusion Hol∗(S2,X )֒→map (S2,X ) k n ∗ n is a homotopy equivalence up to degree k where X ⊆ CPn−1 is the subspace of n pointswithatmostonecoordinatezero. Thecohomologyofthespaceofbasepoint- freeholomorphicmapsHol (S2,S2)wasstudiedbyHavlicek[140,1995]whilethe 1 10 SAMUELBRUCESMITH homotopygroupsofHol (S2,S2)werestudiedbyGuest-Kozlowski-Murayama- k Yamaguchi [121, 1995]. Kallel-Milgram [154, 1997] gave a complete calculation of the homology of Hol∗(T ,CPm) for T an elliptic Riemann surface. The k g g space of real rational functions was recently studied by Kamiyama [157, 2007]. Kallel-Salvatore [155, 2006] applied techniques from string topology to the study of spaces of maps between manifolds. Set H (map(Mm,Nn))=H (map(Mm,Nn)) ∗ ∗+n and similarly for Hol(S2,Nn). When Mm,Nn are closed, compact and orientable, theyprovedH (map(Sm,N))has aringstructurecorrespondingto anintersection ∗ product and H (map(Mm,Nn)) is a module over this ring. They used this struc- ∗ turetocomputeH (map(S2,CPn;ι ))andH (Hol (S2,CPn))withZ -coefficients. ∗ k ∗ k p They also studied H (map(T ,CPn;ι );Z )) for T a compact Riemann surface ∗ g k p g proving, among other results, that these groups are isomorphic for all k when p divides n. 2.1.5. Maps into a Classifying Space and Gauge Groups. LetX beaspaceand Gaconnectedtopologicalgroup. SupposeP: E →X isaprincipalG-bundle. The gauge group G(P) is the topological group of bundle automorphisms of P. The gaugegroupfeaturedinimportantworkofAtiyah-Bott[14, 1983]inmathemat- ical physics. They considered the action of G(P) on the moduli space A of Yang Mills connections on a principal U(n)-bundle P: E → M for M a Riemann manifold. Among other results, they proved H∗(BG(P)) is torsion free and computed its Poincaré series. Their calculation depends on the identity: G(P)≃ Ωmap(X,BG;h), H whereh: X →BGistheclassifyingmapofP,aresultoriginallyduetoGottlieb [112, 1972]. Here X is a finite CW complex. Thus BG(P) ≃ map(X,BG;h). By Bottperiodicity,theloopsanddoubleloopsonBU(n)aretorsionfree. Atiyah-Bott used this fact and Thom’s theory to make their calculations. The classification of gauge groups for fixed X and G up to H-equivalence or, alternately, up to ordinary equivalence is the subject of active research. Gottlieb’s identity implies the homotopy classification problem for map(X,BG) refines the gaugegroupclassificationproblem. Masbaum[199,1991]studiedthehomologyof the components of the space map(X,BSU(2)) for X a 4-dimensionalCW complex obtained by attaching a single 4-cell to a bouquet of 2-spheres. This case includes oriented, simply connected 4-dimensional manifolds. Using a cofibre sequence for X, he obtained, in particular, that the components of map(S4,BSU(2)) represent infinitelymanyhomotopytypes. Usingarelatedanalysis,Sutherland[264,1992] consideredthe classificationofcomponents ofmap(T ,BU(n))forT anorientable g g surface of genus g. He obtained the calculatiuon π2n−1(map(Tg,BU(n);ιk))∼=Zg ⊕Z/d where d=(n−1)!(k,n) thus distinguishing the components corresponding to maps k and l with (k,n) 6= (l,n).Tsukuda [276,2001]classifiedthe homotopytypesrepresentedbythe components ofmap(S4,BSU(2))showingmap(S4,BSU(2);ι )≃map(S4,BSU(2);ι) k l if and only if k = ±l. Kono-Tsukuda [163, 2000] generalized this result from X =S4 to X a simply connected 4-dimensional manifold. As regards the homotopy type of the gauge group, Kono [161, 1991] proved G(P )≃G(P ) ⇐⇒ (12,k)=(12,l) k l